使用具有多个参数和索引的lapply / sapply减少函数内的嵌套循环
Posted
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了使用具有多个参数和索引的lapply / sapply减少函数内的嵌套循环相关的知识,希望对你有一定的参考价值。
我有两个观察向量,我想要计算哪些参数最适合这些观察。因此,我已经定义了似然函数。我正在努力从嵌套for循环到使用sapply。我的目标是返回观察“芽”概率的(50x1)向量。
> bud
[1] 1.72 1.12 1.01 1.14 1.56 1.04 1.00 1.39 1.38
[10] 0.99 1.66 1.55 2.26 0.86 1.45 0.90 1.58 1.09
[19] 0.97 0.98 1.16 1.20 1.17 1.59 1.00 0.92 1.24
[28] 0.98 1.29 1.36 1.80 1.42 1.67 1.13 0.76 0.93
[37] 1.03 1.29 1.09 1.64 1.27 1.14 1.60 1.00 0.91
[46] 1.40 1.15 2.03 1.34 1.37
> deltakere
[1] 8 5 7 6 9 5 5 9 9 6 5 8 8 6 6 5 7 6 7 7 9 6 6 8
[25] 8 6 6 5 9 9 6 7 9 5 5 6 6 8 7 9 8 6 9 8 5 8 8 6
[49] 9 7
当前代码(尽管工作缓慢):
# first equation in my likelihood function
eqOne <- function(i,theta1, theta2, lambda){
vektorN = (1:100)
loopVerdi = 0
if(deltakere[i]>=2){
for (N in deltakere[i]:20000) {
density = ((N)*(N-1)*(pWEI2(bud[i], theta1, theta2))^(N-2) *
pWEI2(bud[i], theta1, theta2, lower.tail = FALSE) *
dWEI2(bud[i], theta1, theta2)) /
(1-(pWEI2(r, theta1, theta2))^N)
ngittN =
dbinom(deltakere[i], size = N,
prob = pWEI2(r, theta1, theta2, lower.tail = FALSE))
sN =
dpois(N, lambda)
loopVerdi = loopVerdi + (density * ngittN * sN)}
return(loopVerdi)}}
# here is the second equation in the likelihood estimate.
eqTwo <- function(theta1, theta2, lambda) {
firstPart <- 1:length(bud)
for (i in 1:length(bud)){
firstPart[i] <- eqOne(i,theta1, theta2, lambda)}
return(firstPart)}
Vector <- eqTwo(1,2,7)
> Vector
[1] 0.071126958 0.174198273 0.125381705 0.202950594
[5] 0.116409283 0.156245797 0.143071864 0.156875579
[9] 0.157670857 0.139149430 0.052085368 0.138248017
[13] 0.001884680 0.073326986 0.161989881 0.103352207
[17] 0.125124465 0.186201434 0.103156346 0.108636007
[21] 0.116075468 0.214339615 0.209921459 0.121433257
[25] 0.084230719 0.102485943 0.216127393 0.135710730
[29] 0.153834163 0.158552633 0.036673600 0.194744214
[33] 0.079498794 0.175511921 0.049722001 0.107646750
[37] 0.159465393 0.198467993 0.169062724 0.089375280
[41] 0.196173239 0.202950594 0.102923165 0.084230719
[45] 0.107521392 0.190136547 0.159243477 0.008274964
[49] 0.158457042 0.210361176
这段代码产生一个(50x1)向量,具有观察“芽”值的概率,这就是我想要的。
这是我尝试从for循环到sapply方法
funDeltakere <- function(i, theta1, theta2, lambda){
function(N, theta1, theta2, lambda){
density = ((N) *(N-1)*(pWEI2(bud[i], theta1, theta2))^(N-2) *
pWEI2(bud[i], theta1, theta2, lower.tail = FALSE) *
dWEI2(bud[i], theta1, theta2)) /
(1-(pWEI2(r, theta1, theta2))^N)
ngittN =
dbinom(deltakere[i], size = N,
prob = pWEI2(r, theta1, theta2, lower.tail = FALSE))
sN = dpois(N, lambda)
return(density * ngittN * sN)}}
eqOne <- function(i,theta1, theta2, lambda){
if(deltakere[i]>=2){
loopVerdi <- sapply(deltakere[i]:20000, funDeltakere(i, theta1, theta2, lambda))
return(loopVerdi)}}
eqTwo <- function(theta1, theta2, lambda) {
firstPart <- 1:length(bud)
for (i in 1:length(bud)){
firstPart[i] <- eqOne(i,theta1, theta2, lambda)
}
return(firstPart)
}
Vector <- eqTwo(1,2,7)
pWEI2(bud [i],theta1,theta2)出错:参数“theta1”缺失,没有默认值
看起来这些参数并没有在我的代码中通过我的第一个函数。任何有关如何确保我的参数通过所有功能的提示将不胜感激!
答案
我设法解决了我的问题。
bud <- runif(10, 1.1,1.55)
deltakere <- round(runif(10,5,9))
r=0.5
foo_inner <- function(i, theta1, theta2, lambda){
function(N){
density = (
(N) *
(N-1) *
(pWEI2(bud[i], theta1, theta2))^(N-2) *
pWEI2(bud[i], theta1, theta2, lower.tail = FALSE) *
dWEI2(bud[i], theta1, theta2)) /
(1-(pWEI2(r, theta1, theta2))^N)
ngittN =
dbinom(deltakere[i], size = N,
prob = pWEI2(r, theta1, theta2, lower.tail = FALSE))
sN =
dpois(N, lambda)
return(density * ngittN * sN)
}}
foo_outer <- function(theta1, theta2, lambda){
function(i){
listeObs <- sapply(deltakere[i]:20000, foo_inner(i, theta1, theta2, lambda))
return(sum(listeObs))
}}
eqThree <- function(theta1, theta2, lambda){
secondPart <- pbsapply(1:length(bud), foo_outer(theta1, theta2, lambda))
LL <- -sum(log(secondPart))
return(LL)
}
result_mle <- mle(minuslogl = eqThree, start=list(theta1 = 1,
theta2 = 2,
lambda = 7),
method="L-BFGS-B", lower=c(0.1,0.1,5),
nobs = length(bud))
以上是关于使用具有多个参数和索引的lapply / sapply减少函数内的嵌套循环的主要内容,如果未能解决你的问题,请参考以下文章